2.3 Anti-Windup and Conditional Integration
2 – 7
2.3
Anti-Windup and Conditional Integration
For better integral performance, the SIM960 features anti-windup
circuitry in the form of conditional integration. The purpose of anti-
windup is to improve the controller’s ability to recover from output
saturation. When the output saturates, the error is likely to be large,
since the process is unable to provide power fast enough to recover
the process output. However, the integrator contribution may not
account for the full amount of the controller output in this case.
Subsequently, the integrator continues to integrate the error until the
integrator output saturates. This “winding up” aspect of integral
control becomes a problem when the process recovers and the error
level passes through zero, because the error must move significantly
beyond zero for the integrator to “unwind” from saturation. In
general, once the controller output is clamped at a limit, nothing is
accomplished by driving it harder into that limit by more integration.
In fact, it only makes it harder to recover from saturation, since the
result is usually large swings back and forth from limit to limit.
There are a variety of anti-windup strategies to mitigate this e
ff
ect. A
simple way to implement anti-windup is to switch o
ff
the integrator
whenever the output saturates. This is not the same as resetting the
integrator (zeroing its output by discharging the feedback capaci-
tance) because the output simply stops moving, but does not go to
zero. It is equivalent to momentarily zeroing the integrator
input
, so
that there is no signal to integrate while the output is saturated.
An improvement to this scheme comes from recognizing that not
all saturation conditions cause unwanted integrator wind-up. For
example, suppose the controller
/
process history were such as to pro-
duce the following conditions:
•
Error signal negative
•
Integrator output finite, not saturated
•
Controller output saturated at the positive limit
Then, the integrator output would be moving in the negative direc-
tion, since its input, the error, is negative. This would not cause
the controller output to be pushed harder into saturation; in fact it
may eventually pull it out of saturation. So stopping the integrator
would hinder the controller’s e
ff
ort to recover the process variable.
The SIM960 uses a technique called “conditional integration:”
Con-
ditional integration only stops the integrator when the polarity of the error
is such as to drive the integrator toward the saturated limit.
SIM960 Analog PID Controller
Содержание SIM960
Страница 4: ...ii Contents SIM960 Analog PID Controller...
